Image features of a splashing drop on a solid surface extracted using a feedforward neural network

This article reports nonintuitive characteristic of a splashing drop on solid surface discovered through extracting image features using feedforward neural network (FNN). Ethanol area-equivalent radius about 1.29 mm was dropped from impact heights ranging 4 cm to 60 (splashing threshold 20 cm) and impacted hydrophilic surface. The images captured when half the were labeled according their outcome, or nonsplashing, used train an FNN. A classification accuracy ≥96% achieved. To extract identified by FNN for classification, weight matrix trained identifying drops visualized. Remarkably, visualization showed that contour height main body impacting as important differentiating between nonsplashing drops, which has not been reported in previous studies. feature found throughout impact, even one three-quarters confirm importance this feature, retrained classify only without checking presence ejected secondary droplets. still ≥82%, confirming is distinguishing drops. Several aspects are analyzed discussed with aim possible mechanism underlying difference